CSBDeep in KNIME – Installation

KNIME Analytics Platform

You can find the current version of the KNIME Analytics Platform on the KNIME website. We will use the version without all extensions. If you choose the other version you can maybe skip some of the following steps if the extension is already installed.

Extensions

To install KNIME Extensions go to Help > Install New Software.... Here you can find new software and install it in your KNIME installation. The extensions are organized in different update sites you need to choose the update site you want to use in the 'Work with:' text field at the top.

Install the following extensions:

• KNIME Python Integration
  • Name: KNIME Python Integration (Labs, supports Python 2 & 3)
  • Update Site: KNIME Analytics Platform 3.4 Update Site - http://update.knime.org/analytics-platform/3.4
• KNIME Image Processing
  • Name: KNIME Image Processing
  • Update Site: KNIME Community Contributions (Nightly) - http://update.knime.org/community-contributions/trunk
• KNIME Image Processing - Python Extensions
  • Name: KNIME Image Processing - Python Extensions
  • Update Site: Stable Community Contributions - http://update.knime.org/community-contributions/3.4

NOTE: Currently, the python integration is buggy. See Current Hacks for solutions. This will be fixed with KNIME 3.5. (If KNIME 3.5.x is the current Version. Feel free to remove this note)

Python

Conda

We will use TensorFlow installed in a conda python environment. Install miniconda3 or anaconda3 as described here. Use miniconda3 if you don't know if you should use anaconda or miniconda.

Python dependencies

Create a new conda environment with the needed dependencies:

conda create -n csbdeep python=3.6 tensorflow pandas jedi

• tensorflow: Needed to load and execute the neural networks. Make sure to use tensorflow-gpu if you have a supported GPU. See the section of your OS in Installing TensorFlow
• pandas: Needed by KNIME
• jedi(optional): Enables python autocompletion in KNIME

Setup the KNIME Python extension

Follow Setting up the KNIME Python extension using the new conda environment csbdeep.

Current Hacks

KNIP Python Extensions: If you get an error, that the input table could not be loaded because the module KNIPImage is missing, just drag a 'Python Script (1⇒1)' node into the workflow and delete it again. The error should not appear again.